Research Of Motion Analysis Technology Based On Human Motion Capture Data

The human motion capture data could preserve the details of the motion, and record the track with the human motion. It has been widely used in many areas such as films, virtual reality and 3D games. With the addition of the motion capture data in the motion capture database, how to make a reasonable and effective analysis of human motion capture data is becoming a hot research topic. The main purpose of motion analysis is to understand and describe the process of motion, so as to manage and reuse the existing motion capture data in the motion capture database.This thesis focuses on behavioral segmentation, motion template extraction and motion recognition based on the motion template. Research works of this thesis mainly include the following aspects:For the behavioral segmentation of human motion capture data, a novel symbolic representation of human motion capture data is presented which represent a high-dimensional motion sequence as a string. The human motion capture data is treated as some high-dimensional data points. This points are clustered by an alternative algorithm based on density, and each cluster is divided into a character. The Behavior String (BS) is produced for the motion data by temporal reverting. The human motion capture data is segmented into distinct behaviors and the cycles of motion are also found by analyzing the BS.For the calculation of motion template, an algorithm for the motion template is presented based on the improved relational features. The relational feature matrix is proposed by the position of each joint. The matrixes from the same behavior are aligned by the Dynamic Time Warping (DTW) method, and the process of time warping is recorded at the same time. The inverse transformation is implemented with the average matrix. Finally, the motion template which could represented the different behaviors are computed by quantization.For the motion recognition of human motion capture data, a novel algorithm is presented which could automatically recognize the sequences from the behavioral segmentation. The sequences after behavioral segmentation are classified into two classes by the trajectory of the Root point from the human skeleton model. The sequences are compared with the motion template by DTW method, and the distance between them are computed. The sequences obtained from the behavioral segmentation are recognized by the DTW matching method.